Climate Dynamics
Influence of air-sea coupling on Indian Ocean Tropical Cyclones
--Manuscript Draft--
Manuscript Number: CLDY-D-17-00779R2
Full Title: Influence of air-sea coupling on Indian Ocean Tropical Cyclones
Article Type: Original Article
Keywords: tropical cyclones
air-sea coupling Indian Ocean cold wake
regional coupled ocean-atmosphere model Corresponding Author: Matthieu Lengaigne, Ph. D.
LOCEAN Paris, FRANCE Corresponding Author Secondary
Information:
Corresponding Author's Institution: LOCEAN Corresponding Author's Secondary
Institution:
First Author: Matthieu Lengaigne, Ph. D.
First Author Secondary Information:
Order of Authors: Matthieu Lengaigne, Ph. D.
S Neetu
Guillaume Samson Jerôme Vialard KS Krishnamohan Sebastien Masson Swen Jullien I Suresh
Christophe Menkes Order of Authors Secondary Information:
Funding Information: IFCPAR
(4907-1) Dr. Matthieu Lengaigne
Abstract: This paper assesses the impact of air-sea coupling on Indian Ocean Tropical Cyclones (TCs) by comparing a 20-year long simulation of a ¼° regional coupled ocean-
atmosphere model with a twin experiment, where the atmospheric component is forced by sea surface temperature from the coupled simulation. The coupled simulation reproduces the observed spatio-temporal TCs distributions and TC-induced surface cooling reasonably well, but overestimates the number of TCs. Air-sea coupling does not affect the cyclogenesis spatial distribution but reduces the number of TCs by ~20%
and yields a better-resolved bimodal seasonal distribution in the northern hemisphere.
Coupling also affects intensity distribution, inducing a four-fold decrease in the proportion of intense TCs (stronger than Cat-1). Air-sea coupling damps TC growth through a reduction of the inner-core upward enthalpy fluxes due to the TC-induced cooling. This reduction is particularly large for the most intense TCs of the northern Indian Ocean (up to 250W.m-2), due to higher ambient surface temperatures and larger TC-induced cooling there. The negative feedback of air-sea coupling on strongest TCs is mainly associated with slow-moving storms. Sensitivity experiments using a different convective parameterization yield qualitatively similar results, with a
larger (~65%) reduction in the number of TCs. Because of their relatively coarse resolution (¼°), both set of experiments however fail to reproduce the strongest observed TCs (Cat-4-5). Further studies with finer resolution models capable of simulating the strongest TCs in the Bay of Bengal will be needed to assess the expectedly large impact of air-sea coupling on those deadly TCs.
Reviewer original comment – Authors answer – Actions taken in the text
Reply to the review comments Reviewer #1:
Reviewer comments:
The paper entitled "Influence of air-sea coupling on Indian Ocean Tropical Cyclones"
by Lengaigne et al. investigated roles of air-sea coupling and subsurface ocean to SST cooling during TC passage. From regional air-sea coupled model experiments, the authors suggest that SST cooling (cold wake) causes TC reduction particular in intense TCs (Fig.5 and 6) and cold wake is larger over the northern IO than southern IO (Fig.8 and 9). Impact of subsurface ocean on intense TC over the northern IO is also summarized in Fig.10 and 11. I think that this paper is well written based on suitable analysis, and therefore I would like to accept this paper after minor revision.
We thank the reviewer for his positive views on our manuscript.
Comments:
- Our previous work (Ogata et al. 2016) also pointed out the importance of subsurface water distribution over the tropical Indian Ocean (although our result was more significant over the southern IO than northern IO) on intense/extreme TC distribution.
Could you please discuss similarity and different between Ogata et al. (2016) and this paper? Particularly, it is interesting that this paper shows weakening of intense TCs and air-sea coupling effect is weak in the southern IO. This result is different from Ogata et al. (2016) that the thermocline ridge over the southwestern IO causes significant air-sea coupling effect in the southern IO.
We thank the reviewer for this reference. Results from this paper are now discussed in the introduction (p4 first paragraph and p6 second paragraph) and compared to our results at the end of Section 3 (p15 first paragraph). The following discussion has been added: “In agreement with Ogata et al. (2016), we do not find any significant change in the location of the region of strongest TCs when air-sea coupling is accounted for, neither in the northern nor in the southern IO (not shown). However, while Ogata et al. (2016) found a considerably larger decrease of the number of strong TCs in the southern than in the northern IO, our results indicate a more symmetrical decrease on both sides of the equator.”
- In Fig.4, is it possible to quantify (or detect) which component (potential intensity, vorticity, and vertical shear) is important for bi-modal seasonality? For example, converting units of Fig.4b-4e and 4g-4j to GPI may also give a useful information.
The assessment of the respective influence of each term on the GPI seasonal evolution is not straightforward. As this index is the product (and not the sum) of different terms, these terms interact non-linearly to shape the GPI evolution. Menkes et al.
(2011) applied an efficient method to quantify the relative roles of each of the large- scale environmental factors in causing the bimodal distribution of the GPI in the northern IO. This method, based on Camargo et al. (2007), used a first order Taylor expansion of the logarithm of the GPI equation to express the GPI variations as the sum of different terms (see Menkes et al. 2011 for details). Menkes et al. (2011) found using this method that the combined effect of increased vertical wind shear and decreased maximum potential intensity overcome the increased relative humidity during the summer monsoon. To ascertain the robustness of their results, we applied
Authors Click here to download Authors' response to reviewers'
comments Lengaigne Cyc Coupling reply final.docx
Reviewer original comment – Authors answer – Actions taken in the text
the same diagnostics on our dataset and results are presented on Figure R1 (to be compared with the bottom left panel of Figure 7 from Menkes et al. 2011).
Figure R1. Evolution of the GPI seasonal anomalies (dGPI) and relative contributions of each term included in the GPI (Vorticity, Shear, Humidity, Vpot) to this seasonal evolution as derived from the Taylor expansion method described in Menkes et al. (2011) and their sum in the northern IO (SUM) for observations. See Menkes et al. (2011) for details.
In line with Menkes et al. (2011), Figure R1 shows that the GPI decrease in summer arises largely because of the increased vertical wind shear and more modestly from decreased maximum potential intensity, which allow overcoming the increased relative humidity in that season. We however decided not to include this analysis in the revised manuscript for several reasons. First, our results are very consistent with the analysis of the terms driving the seasonal evolution of the GPI of Menkes et al.
(2011). Second, explaining the decomposition method will unnecessarily lengthen the manuscript. And third and most importantly, Menkes et al. (2011) also demonstrated by comparing several cyclogenesis indices that, if the shear increase is always the dominant factor in driving this summer seasonal decrease of these indices, the quantitative contribution from the other terms vary from one index to another. We however mention in the last paragraph of page 13 that: “Our results are consistent with Menkes et al. (2001) who attributed this bimodal distribution to the dominant effect of increased vertical wind shear that overcome the opposite effect of increased relative humidity during the summer monsoon (Figures 4f–j), preventing the formation of tropical storms during that season.”
- In Fig.5, I understood that air-sea coupling is less important for weak TCs (determined mostly by SST) while TC density significantly decreases in intense TCs by air-sea coupling induced cold wake. Is it right?
Yes, this is correct.
Reference:
Ogata, T., Mizuta, R., Adachi, Y., Murakami, H., & Ose, T. (2016). Atmosphere-
ocean coupling effect on intense tropical cyclone distribution and its future change
with 60 km-AOGCM. Scientific reports, 6, 29800.
Reviewer original comment – Authors answer – Actions taken in the text
Reviewer #2:
Review of "Influence of air-sea coupling on Indian Ocean tropical cyclones" by Lengaigne et al.
This study uses a 0.25-deg couple model to simulate Indian Ocean tropical cyclones and estimate the impact of SST coupling on TC tracks and intensities. The results are mostly what would be expected: air-sea coupling reduces TC frequency and intensity, especially for the strongest storms. The reductions are due mainly to the presence of storm-induced cold wakes in the coupled run, which reduce the upward energy transfer to the TCs. The effect is strongest for the most slowly moving storms, which spend the longest time over their cold wakes, and is strongest in the northern Indian Ocean (Bay of Bengal), where the upper-ocean temperature stratification is stronger.
This is an interesting and well written manuscript that presents new and important results. The 0.25-deg resolution model that is used is not ideal because it can't simulate the strongest TCs, and it produces significantly more TCs compared to observations (about 4 times as many for the coupled run). However, the model does compare well to observations in other aspects such as spatial distribution of TCs and seasonal distribution of TC number in each basin. The model is also capable of producing TCs with well-defined eyes and reasonable radii of maximum winds, suggesting that the horizontal resolution is adequate for examining large-scale statistics of simulated TCs. I do not see it as a major limitation that the model produces too many TCs (except for one minor point that I include below). The fact that the model does not produce enough strong TCs could mean that the results err toward the conservative side, since stronger storms tend to produce more mixing and SST cooling, with a stronger negative feedback. It is good to see that the authors experimented with two different convective parameterizations and that the overall features are similar despite one method failing to produce any strong storms. It's valuable to know the sensitivity of modeled TCs to this parameterization.
I have some mostly minor points for the authors to consider, after which the manuscript should be suitable for publication.
We thank the reviewer for his positive views on our manuscript.
Because there are so many more TCs in the model compared to observations, there may be a spurious negative feedback induced by prior storms' cold wakes (Balaguru et al., GRL, 2014, doi:10.1002/2014GL061489). I would expect this might be most important in the Bay of Bengal due to its small area. Can you quantify this effect, or estimate the chance of one TC encountering the wake of a previous TC?
This is a good point. We hence calculated the probability for each TC location to be within 400km and 14 days-lag of a previous TC, for both the northern and southern hemispheres. The 400km and 14 days have been chosen here to broadly match criteria in Balaguru et al. (2014). Our results indicate that this probability is low in observations for both northern and southern IO (7.4 and 5.1% respectively). Although larger for the coupled simulation because of the larger number of TCs (18.6 and 7.5%
respectively), it still small enough to conclude that the effect of spurious negative
feedback induced by previous cold wake to be of second order compared to the direct
Reviewer original comment – Authors answer – Actions taken in the text
negative feedback induced by the TC itself. This probability decreases further when less conservative criteria are used. The cooling 400km away from a TC is indeed very weak (3 to 4 times less than the maximum cooling, see Figure 7b from Vincent et al.
2012 and Figure 8 and 9 from the present study) and is unlikely to affect significantly another TC passing several days after. Using a more realistic 200km criterion, this probability decreases to 1% in observations and 8% in the model.
p. 4, first paragraph: Others have shown the importance of an accuate representation of ocean stratification on basin scales and might be worth citing (Balaguru et al., GRL, 2015, doi:10.1002/2015GL064822 and Miyamoto et al., GRL, 2017, doi:10.1002/2017GL073670).
Thanks for these references. They are now discussed in the introductory section (first paragraph of p4) of the revised manuscript.
p. 5, line 11: Delete 'the' after 'improves' Corrected.
p. 9, lines 9-19: I'm curious how well the wake removal algorithm works. It would be easy enough to test by doing an experiment in which random synthetic "tracks" are removed from the coupled model output and then filled using your technique. The filled values could then be compared to the actual values. Its unclear whether such a validation was done previously in another paper, and if so it should be mentioned.
The filling technique used in our manuscript and Samson et al. (2014) was also used in Jullien et al. (2014), as already mentioned in the manuscript. Figure 2 illustrates that this technique efficiently removes the TC cold wake with affecting the SST away from the TC track. Similar consistency checks have been performed for numerous other TC cases, with similar results. In addition, due to the chaotic nature of the atmosphere, TCs occur at different location in the forced and coupled simulation, making it very unlikely that they will be heavily influenced by other TCs cold wakes, even if they were not perfectly “filled”. The composite of figure 9 illustrates this:
while there is a very clear cold wake signature in the coupled model, this signature is absent from the forced model.
p. 12, lines 21 and 23: I think you mean to refer to Figure 3, not Figure 2.
We actually referred to Figure 1. Thanks for noticing. This has been changed.
p. 14, line 17: Change 'simulates' to 'simulate' Corrected.
p. 19, line 19: Change 'to' to 'on' Corrected.
p. 20, line 14: Change 'twice larger' to 'twice as large'
Corrected.
Reviewer original comment – Authors answer – Actions taken in the text
Reviewer #3:
Review of "Influence of air-sea coupling on Indian Ocean tropical cyclones" by Lengaigne et al.
Most present-day models do not have adequate skill in cyclone intensity simulation or forecasts, partly due to poor representation of the physics of air-sea interaction. This is a scientific problem of great applied interest.
The authors have made a significant contribution, in the modeling context, to the role of air-sea interaction in Indian Ocean tropical cyclone numbers, their seasonality, spatial distribution, and intensification. They analyse 20-year simulations from a ¼ degree resolution regional coupled model; compare the results with a standalone atmosphere model forced by SST from the coupled model, and thus elicit the influence of cyclone-induced SST cooling on intensification of tropical cyclones in the north and south Indian Ocean.
The present work draws upon previous studies of tropical cyclones published by one or more of the authors with their collaborators, and it can be seen as part of a sequence of studies. The authors use the NEMO-WRF coupled model, and the analysis employs indices of intensity and inhibition developed earlier, as well as proxies of background upper ocean stratification. The ¼ degree model generates too many weak systems relative to observations, and does not simulate the most intense (category 3 or higher) storms.
The main conclusions are: Air-sea coupling does not change the spatial distribution of Indian Ocean cyclogenesis, but reduces the overall number of TCs by 20%. Coupling reduces the number of TCs in the north Indian Ocean in the summer monsoon season, bringing the model seasonality closer to observations. Reduction in enthalpy flux and TC intensity due to coupling is more pronounced in the north Indian Ocean than in the south. These results are robust under a change of convective parameterization scheme; the reduction in overall TC numbers is much larger when the Kain-Fritsch parameterization is employed.
The model simulations, experiment design and analysis are all sound, model shortcomings are explicitly discussed, and the simulations are compared with observations at each step. The results are placed in the context of the handful of previous studies that use similar long model simulations focused on Pacific and Atlantic TC’s. The paper is well organized, although the writing does need more clarity in several places.
Recommendation: Accept with minor revisions.
We thank the reviewer for his positive views on our manuscript.
The present study generally clubs the cyclone seasons together while examining
cyclone inhibition. The pre-summer monsoon and post-summer monsoon temperature
and salinity stratification in the north Indian Ocean have significant differences, as
Reviewer original comment – Authors answer – Actions taken in the text
shown in previous observational studies. Although seasonal differences are not the focus of this study, the authors may mention this explicitly in the Discussion Section.
We did not want to include a discussion on this aspect as we felt the paper was long enough.
Many sentences in the text need to be rewritten for greater clarity. I mention a few examples later; first, I suggest changing the Abstract slightly to:
This paper assesses the impact of air-sea coupling on Indian Ocean Tropical Cyclones (TCs) by comparing a 20-year long simulation of a ¼° regional coupled ocean- atmosphere model with a twin experiment, where the atmospheric model is forced by sea surface temperature (SST) from the coupled simulation. The coupled simulation reproduces the observed spatio-temporal TC distribution and TC-induced surface cooling reasonably well, but overestimates the number of TCs. Air-sea coupling does not affect the spatial distribution of cyclogenesis, but reduces the number of TCs by
~20% and yields a better-resolved bimodal seasonal distribution of TCs in the northern hemisphere. Coupling also affects intensity distribution, inducing a four-fold decrease in the proportion of intense TCs (stronger than category-2). Air-sea coupling damps TC intensity through a reduction of the inner-core upward enthalpy flux due to TC-induced SST cooling. This reduction is particularly large for the most intense TCs of the northern Indian Ocean (up to 250 Wm-2), due to higher ambient surface temperatures and larger TC-induced SST cooling. The negative feedback of air-sea coupling on the strongest TCs is associated mainly with slow-moving storms.
Sensitivity experiments using a different atmospheric convective parameterization yield qualitatively similar results, with a larger (~65%) reduction in the number of TCs. Because of the relatively coarse model resolution (¼°), however, both sets of experiments fail to reproduce the strongest observed TCs (category 4-5). Further studies with finer resolution models capable of simulating category 4-5 TCs will be needed to assess the full impact of air-sea coupling, particularly on the deadly cyclones in the Bay of Bengal.
We implemented most of the changes suggested by the reviewer in the abstract.
Specific examples of sentences/lines that need more clarity or grammar checks, by page and line number; there are more.
We went throughout the manuscript again and did our best to clarify the text and correct grammar checks.
• Page 2, lines 13-14 –“This air-sea coupling ...cold wake they induce”.
Changed.
• Page 4, line 13 – “idealized case studies”.
Corrected.
• Page 4, line 26 – “Only a handfull of studies...”.
We believe the right spelling is « handful » and not « handfull ».
• Page 6, line 6 – “observation-based studies”.
Corrected.
Reviewer original comment – Authors answer – Actions taken in the text
• Page 6, line 14 – “favouring an enhanced cooling”.
Corrected.
• Page 9, line 9 – space after “Jullien et al. (2014),”
Done.
• Page 9, line 12 – Extra space.
Done.
• Page 10, lines 11-12 – “3, 6 and 9 radii”.
We don’t understand the suggested correction.
• Page 11, line 7 – Extra spaces.
Done.
• Page 11, line 8 – “Vshear”.
Corrected.
• Page 12, lines 6-7.
We don’t understand this comment.
• Page 12, line 21 – Change (Figure 2bc) to (Figure 1b, c).
Thanks, this has been corrected.
• Page 12, line 23 – Change (Figure 2d) to (Figure 1d).
Corrected.
• Page 14, line 1 – Change (Figure 3bf) to (Figure 3b, f) – many more like this.
Corrected everywhere.
• Page 14, line 17 – “does not simulates”
We don’t think this is the correct spelling.
• Page 17, lines 16 – “In the southern hemisphere main ...”.
Done.
• Page 18, line 6 – add “wind stress, Ekman pumping and westward moving planetary/Rossby waves” ? Rossby waves may be important in both hemispheres Yes, done.
• Page 18, lines 10-18 – a little more clarity possible ? We tried to do so.
• Page 19, line 8 – use “radius of maximum wind’ instead of RMW.
Changed.
• Page 20, lines 23 – Figure 10.
Changed.
Reviewer original comment – Authors answer – Actions taken in the text
• Page 23, line 5 – “adjusted towards”
Corrected.
• Page 35, lines 5-6 – "TC track from the is shown…” sentence is incomplete.
Corrected.
• Page 36 – Modify Figure 3 – left and right columns should be interchanged.
We kept it as it was.
• Page 44, lines 5-7 – blank space before brackets.
Changed.
Unité mixte de recherches 7159 UPMC, CNRS, MNHN, IRD Matthieu Lengaigne
LOCEAN – Université Pierre et Marie Curie Case 100 – 4, place Jussieu
75252 Paris Cedex 05 - France matlaldp@gmail.com
About: Revised version of CLDY-D-17-00779, by Lengaigne et al.
Dear editor,
We thank the Editor and Reviewers for providing valuable comments on our manuscript. Both reviewers suggested minor edits, which we all implemented. We also provide a separate document including point-by-point answers to reviewers. We believe that all the queries raised by the Reviewers have been addressed in the revised manuscript.
We hope that this revised manuscript will meet the Climate Dynamics requirements.
Sincerely, M. Lengaigne
Cover letter
1
Influence of air-sea coupling on Indian Ocean Tropical Cyclones
2 3 4
Lengaigne, M.
1,2, S. Neetu
3,
G. Samson
4, J. Vialard
1, K.S Krishnamohan
1, S. Masson
1, S. Jullien
5, I.
5
Suresh
3, C. E. Menkes
1,66
7 8
1- Sorbonne Universités (UPMC, Univ Paris 06)-CNRS-IRD-MNHN, LOCEAN Laboratory, IPSL, Paris, France
9
2- Indo-French Cell for Water Sciences, IISc-NIO-IITM–IRD Joint International Laboratory, NIO, Goa, India
10
3- CSIR-National Institute of Oceanography, Goa, India
11
4- Mercator Océan, Ramonville-Saint-Agne, France
12
5- Laboratoire d'Océanographie Physique et Spatiale (LOPS), Univ. Brest, CNRS, IRD, Ifremer, IUEM, 29280 Plouzané, France
13
6- Centre IRD, Nouméa, New Caledonia
14 15
Revised for Climate Dynamics 16
5
thFebruary 2018 17
18 19
Corresponding author:
20
M. Lengaigne 21
National Institute of Oceanography 22
403004 Dona Paula 23
Goa, India 24
Email: lengaign@locean-ipsl.upmc.fr 25
26
Manuscript Click here to download Manuscript Lengaigne Cyc Coupling
text revised final.docx
Click here to view linked References
1
Abstract
2
This paper assesses the impact of air-sea coupling on Indian Ocean Tropical Cyclones (TCs) by 3
comparing a 20-year long simulation of a ¼° regional coupled ocean-atmosphere model with a twin 4
experiment, where the atmospheric component is forced by sea surface temperature from the coupled 5
simulation. The coupled simulation reproduces the observed spatio-temporal TCs distributions and TC- 6
induced surface cooling reasonably well, but overestimates the number of TCs. Air-sea coupling does not 7
affect the cyclogenesis spatial distribution but reduces the number of TCs by ~20% and yields a better- 8
resolved bimodal seasonal distribution in the northern hemisphere. Coupling also affects intensity 9
distribution, inducing a four-fold decrease in the proportion of intense TCs (stronger than Cat-12). Air-sea 10
coupling damps TC growth through a reduction of the inner-core upward enthalpy fluxes due to the TC- 11
induced cooling. This reduction is particularly large for the most intense TCs of the northern Indian Ocean 12
(up to 250W.m
-2), due to higher ambient surface temperatures and larger TC-induced cooling there. The 13
negative feedback of air-sea coupling on strongest TCs is mainly associated with slowly-moving storms.
14
Sensitivity experiments using a different convective parameterization yield qualitatively similar results, 15
with a larger (~65%) reduction in the number of TCs. Because of their relatively coarse resolution (¼°), 16
both set of experiments however fail to reproduce the strongest observed TCs (Cat-4-5). Further studies 17
with finer resolution models capable of simulating the strongest TCs in the Bay of Bengal will be needed to 18
assess the expectedly large impact of air-sea coupling on those deadly TCs.
19
20
1. Introduction
1 2
Tropical cyclones (TCs) track prediction has dramatically improved over the last decades, yet intensity 3
forecasts improvements are more limited (DeMaria et al. 2014). Internal dynamics, environmental forcing, 4
and interactions with the ocean are generally identified as important elements affecting TC intensity 5
evolution (Wang and Wu 2004). Various atmospheric large-scale conditions influence TC intensification, 6
such as strong vertical wind shear that increases inner-core static stability due to the vortex tilting (DeMaria 7
1996) or a dry mid-troposphere, which inhibits convective buoyancy (Emanuel et al. 2004). As TC 8
primarily draw their energy from evaporation at the surface of the ocean (e.g. Riehl 1950; Emmanuel 9
1986), the enthalpy fluxes at the air–sea interface also play an essential role on TCs intensification (e.g.
10
Emmanuel 1999). While the ocean provides the necessary thermal energy for TCs through moist surface 11
enthalpy flux, TC intensity is also sensitive to the local evolution of Sea Surface Temperature (SST) under 12
the storm eye (Schade 2000). The kinetic energy dissipated due to the strong wind stress at the air-sea 13
interface (Emanuel 2003) results in a significant SST cooling under the TC, largely through vertical mixing 14
(e.g. Price 1981; Jullien et al. 2012; Vincent et al. 2012a). This TC-induced cooling acts to limit the TC 15
intensification (Cione and Uhlhorn 2003). The fundamental role of air-sea heat exchanges in the TC 16
intensification paradigms hence emphasizes the need for quantifying and understanding the feedback of this 17
TC-induced cooling on TC characteristics.
18 19
Through a statistical analysis of 23 TCs in the Atlantic, Cione and Ulhorn (2003) provided observational 20
evidences to support the key role played by air-sea coupled feedbacks on TCs intensification: in this limited 21
dataset, TCs inducing large SST cooling indeed generally experience a weaker upward surface enthalpy 22
flux (up to 40% reduction for a ~1°C change) and a weaker TC intensification rate. The levelling-off of 23
TC-induced cooling with increasing TC intensity (from Cat-2) was further interpreted as an indirect 24
evidence of the negative ocean feedback onto TC intensification (Lloyd and Vecchi 2011). The faster TC 25
translation speed for increasing TC intensity was also interpreted as indirect evidence of air-sea coupling 26
(Mei et al. 2012): faster-moving TCs spend less time over a given ocean location and consequently imprint
27
a weaker SST cooling, hence favoring a stronger TC intensification. Including an upper ocean parameter in 1
statistical intensity forecasting reduces TC intensity forecast errors by ~5% on average (DeMaria et al.
2
2005; Mainelli et al. 2008). Similarly, accounting for ocean stratification allows increasing the explained 3
variance of TC intensification accounted by the maximum potential intensity theory (Balaguru et al. 2015;
4
Miyamoto et al. 2017). This indicates that the oceanic subsurface stratification also influences the cooling 5
under the TC, with more pronounced cooling in regions of shallow thermocline, such as cyclonic eddies, 6
that may feedback more on the TC. This is supported by the numerous observational case studies that 7
reported a rapid change of the intensification rate of TCs passing over oceanic eddies (e.g., Shay et al.
8
2000; Lin et al. 2005). Similarly, at the large-scale, the ocean negative feedback on the TCs varies due to 9
contrasted oceanic stratifications in different basins (e.g. Vincent et al. 2014).
10 11
A large body of literature using coupled atmosphere–ocean models further demonstrated that North 12
Atlantic and Pacific TCs intensity forecasts were significantly improved when the TC-induced ocean 13
feedback was accounted for (some cyclones were otherwise over-intensifying; e.g., Bender and Ginis 2000;
14
Hong et al. 2000; Lin et al. 2005; Ito et al. 2015). Other coupled modelling idealized case studies confirmed 15
that the TC-induced cooling limits TC intensification (e.g. Bender et al. 1993; Schade and Emmanuel 1999;
16
Zhu et al. 2004; Chen et al. 2010; Liu et al. 2011; Ma et al. 2013; Halliwell et al. 2015) but could also 17
impact the TC size (Chen et al. 2010; Ma et al. 2013) and asymmetrical structures (Zhu et al. 2004; Chen et 18
al. 2010). TC tracks are on the other hand rather insensitive to air-sea coupling (Zhu et al. 2004; Chen et al.
19
2010; Liu et al. 2011). Finally, it has become increasingly evident that the oceanic component of these 20
models needs to include three-dimensional processes to correctly simulate the TC-induced upper ocean 21
response, especially for slow-moving TCs (Yablonsky and Ginis 2009; Halliwell et al. 2011).
22 23
The studies discussed above assessed the impact of air-sea coupling on TCs from case studies of specific 24
TCs, using short-term coupled model integrations. This is enlightening for the processes involved, but does 25
not allow estimating the climatological effect of air-sea coupling on TCs statistics (count, distribution,
26
intensity, etc…). Only a handful of studies using high-resolution climate models, able to explicitly simulate 1
realistic TCs, did investigate this climatological effect by comparing long-term (~20 years) coupled and 2
uncoupled model integrations at global scale (Ogata et al. 2016) and regionally for the Pacific (Jullien et al.
3
2014) and Atlantic TCs-prone regions (Ogata et al. 2015; Zarzycki 2016). These studies indicate that air- 4
sea coupling reduce the number of TCs numbers by 5 to 9% in the North Atlantic and Pacific (Zarzycki 5
2016) and by 10% in the South Pacific (Jullien et al. 2014). Air-sea coupling in particular reduces 6
cyclogenesis in the Coral Sea, where a shallow thermocline promotes large TC-induced cooling, leading to 7
a more realistic cyclogenesis pattern in the South Pacific (Jullien et al. 2014). In contrast, cyclogenesis 8
patterns were not considerably affected by air-sea coupling in the north Pacific and Atlantic (Zarzycki 9
2016). These studies also indicate that accounting for air-sea coupling yields a strong reduction of the 10
number most intense TCs simulated in those models. This reduction is not spatially uniform in the northern 11
hemisphere (Ogata et al. 2015, 2016) resulting in an equatorward shift of the maximum of intense TC 12
distribution in the North Atlantic and North Pacific that improves the representation of their climatological 13
distribution and north-south asymmetry. The shallow northwest Pacific and Atlantic thermocline indeed 14
promotes intense TC-induced cooling, hence limiting TC intensification and reducing the number of most 15
intense TCs there. The impact of air-sea coupling operates through a reduction of the enthalpy flux fuelling 16
the TC in response to the TC induced cooling. The enthalpy flux reduction is ~20% when averaged over all 17
TCs (Jullien et al. 2014; Zarzycki 2016) but reaches up to ~26% for hurricanes (>33 m.s
-1) in north Pacific 18
and Atlantic (Zarzycki 2016).
19 20
Except for the global study of Ogata et al. (2016), all short and long-term coupled modelling studies 21
mentioned above have been performed either in the Atlantic or Pacific basins. The Indian Ocean (IO) is 22
however home to about 20% of the global TC activity. TCs in the northern IO mainly occur in the western 23
and central part of the Bay of Bengal (BoB; Fig. 1a). The strong vertical shear during the southwest 24
monsoon inhibits cyclones, leading to a bimodal seasonal distribution, with preferential occurrences during 25
the pre- and post-monsoon (Menkes et al. 2011, Li et al. 2013). Although northern IO just accounts for 7%
26
of TCs worldwide, those TCs have catastrophic impacts, with 14 of the 20 deadliest world TCs in recent 1
history having occurred in the BoB (Longshore 2008). TCs in the southern IO occur over an elongated 2
band centred on 15°S from November to April (Fig. 1a), with enhanced TC occurrence over the 3
southwestern IO around the Mauritius, La Reunion and Madagascar islands (Mavume et al. 2009). Only a 4
few studies reported a TC-induced upper ocean cooling in the northern (e.g., Sengupta et al. 2008;
5
McPhaden et al. 2009; Neetu et al. 2012; Girishkumar et al. 2014) and southern IO (Vialard et al. 2009).
6
Even fewer observation-based studies discussed the potential feedback of this oceanic response on IO TC 7
intensification, and only for the case of the BoB (Ali et al. 2007; Lin et al. 2009; Yu and McPhaden 2011).
8
Analyses of moored buoys data for instance indicated that a combined warm subsurface layer and strong 9
salinity stratification played an essential role in the rapid intensification of the Nargis TC in the BoB (e.g.
10
Lin et al. 2009; Yu and McPhaden 2011).
11 12
Very different upper ocean thermohaline structures in the two IO TC-prone regions may result in a different 13
sensitivity of TCs to ocean-atmosphere coupling. The BoB is indeed characterized by a strong haline 14
stratification that may limit the amplitude of TC-induced cooling and promote TC intensification (Sengupta 15
et al. 2008; Neetu et al. 2012). In stark contrast, the cyclogenesis region in the southwestern IO is one of 16
the rare oceanic regions where warm SSTs coexist with a shallow thermocline ridge (Vialard et al. 2009), 17
hence potentially favouring an enhanced cooling below the storm and a strengthened negative oceanic 18
feedback at the early stages of the storm intensification (Xie et al. 2002). Ogata et al. (2016) results suggest 19
that these differences may be responsible for a stronger reduction of intense TCs in the southwestern IO 20
than in the BoB but their global perspective and their focus on intense TCs did not allow a detailed 21
quantification of the influence of air-sea coupling on TC characteristics in the IO. To our knowledge, this 22
study is the first one to address this question using a regional coupled model simulation over a long period.
23 24
We use a regional IO, ~25-km resolution coupled ocean-atmosphere model (Samson et al. 2014) to assess 25
the negative feedback of air-sea coupling on the TC amplitude in this basin. This model simulates realistic 26
spatial and seasonal distributions of IO tropical cyclones, as well as their variations in association with the
27
leading modes of IO interannual climate variability (Samson et al. 2014). We will compare TC statistics 1
from two twenty year-long experiments (a reference coupled experiment and forced atmospheric 2
experiment, i.e. with no air-sea coupling) to provide a reliable statistical assessment of the impact of air-sea 3
coupling on IO TCs. This approach has already been successfully used to study the effect of air-sea 4
coupling on the south Pacific TC climatology (Jullien et al. 2014). The rest of the paper is organised as 5
follows. Section 2 describes the observed datasets and the modelling framework. In section 3, we first 6
validate the climatological TC spatio-temporal distribution and assess the impact of air-sea coupling on IO 7
TC characteristics. Section 4 will validate the TC-induced cooling in the coupled simulation and discuss the 8
mechanisms of the negative air-sea coupling feedback on TCs. Section 5 discusses the robustness of results 9
presented in the previous sections when choosing a different convective parameterization in the 10
atmospheric model. The final section provides a summary and a discussion of our results.
11 12 13
2. Datasets and methods
14 15
2.1. The NOW regional coupled model
16 17
In this study, we use an IO configuration of the NOW (NEMO-OASIS3-WRF) regional coupled model.
18
NOW couples the NEMO (Nucleus for European Modelling of the Ocean) ocean general circulation model 19
(Madec et al. 2008) to the WRF (Weather Research and Forecasting Model) regional atmospheric model 20
(Skamarock and Klemp 2008) through the OASIS3 coupler (Valcke 2013). This configuration has been 21
extensively described and validated in Samson et al. (2014). We therefore only provide a brief summary of 22
this configuration in the following.
23 24
The ocean model parameterizations include a turbulent kinetic energy scheme for vertical mixing (Blanke 25
and Delecluse 1993), bi-laplacian viscosity and an iso-neutral laplacian diffusivity for subgrid-scale mixing 26
(Lengaigne et al. 2003). Atmospheric model physics include the WRF single-moment six-class
27
microphysics scheme (Hong and Lim 2006), the Goddard shortwave radiation scheme (Chou and Suarez 1
1999), the Rapid Radiation Transfer Model for longwave radiation (Mlawer et al. 1997), the Yonsei 2
University planetary boundary layer (Noh et al. 2003) and the four-layer Noah land surface model (Chen et 3
al. 1996). It also includes the updated Kain-Fritsch (KF) atmospheric convective scheme (Kain 2004), a 4
mass-flux convergence scheme that allows shallow convection, includes a minimum entrainment rate to 5
suppress widespread convection in marginally unstable relatively dry environments, and an updated 6
downdraft formulation. This scheme provides the best simulations of TC intensity and track prediction in 7
the northern IO (Srinivas et al. 2013), with higher convective warming and stronger vertical motions 8
relative to the other tested cumulus schemes. As discussed in Samson et al. (2014), the KF scheme allows 9
simulating stronger TCs (up to 56 m.s
-1) than the Betts-Miller-Janjic (BMJ) moist convective adjustment 10
scheme (up to 45 m.s
-1, Janjic 1994) but leads to a large overestimation of the number of IO TCs. Deep 11
convection is an essential mechanism for TCs, and the results presented here may be sensitive to the 12
convective scheme. The sensitivity of our results to the choice of convective scheme will thus be addressed 13
in section 5 by comparing results obtained using KF with those obtained using the BMJ convective scheme.
14
The wind stress drag coefficient parameterization over the ocean is based on the work of Donelan et al.
15
(2004).
16 17
This configuration covers the IO region [25.5°E-142.5°E, 34.5°S-26°N], with the oceanic and atmospheric 18
component sharing the same 1/4
ohorizontal grid. The ocean component has 46 vertical levels, with an 19
enhanced 5 m resolution in the upper ocean. The atmospheric component has 28 sigma vertical levels, with 20
a higher resolution of 30 m near the surface. NEMO lateral boundary conditions are prescribed from a 1/4
o21
resolution global ocean model forced by Drakkar forcing dataset (Brodeau et al. 2010). Those of WRF are 22
taken from 6 hourly ERA-Interim reanalysis (Dee et al. 2011). The 1st of January 1989 initial condition is 23
provided from ERA-Interim for the atmosphere and from the ¼° DRAKKAR simulation mentioned above 24
for the ocean. A 21 year coupled simulation is performed with this setup using 1989–2009 lateral boundary 25
conditions, and will be referred to as KF-CPL (for Kain-Fritsch coupled). The first year of this experiment
26
is discarded, and this simulation is hence analysed over a 20-years period.
1 2
Samson et al. (2014) demonstrated that this configuration is able to capture the main features of the Indian 3
Ocean climate. At seasonal timescales, it reproduces the seasonal rainfall distribution and the northward 4
seasonal migration of monsoon rainfall over the Indian subcontinent. It also captures the observed 5
interannual variability associated with the Indian Ocean Dipole (Saji et al. 1999; Webster et al. 1999) and 6
El Niño Southern Oscillation (e.g. Xie et al. 2009). More importantly for our study, its relatively high 7
horizontal resolution allows to explicitly simulate TCs, with realistic cyclogenesis, track density patterns 8
and a realistic seasonal cycle, including the observed bimodal distribution in the northern IO (Samson et al.
9
2014). The seasonal evolution of the large-scale atmospheric parameters involved in TC genesis is also 10
properly captured.
11 12
2.2. Sensitivity experiments
13 14
We follow a similar strategy to Jullien et al. (2014) to isolate the influence of air-sea coupling on TCs. We 15
perform a twin uncoupled atmospheric simulation (referred to as KF-FOR) using the same WRF 16
atmospheric configuration, lateral boundary conditions and SST fields from the KF-CPL simulation, from 17
which TC-induced cold wakes have been suppressed. Cold wake signals are removed by masking the 18
coupled model SST within 3° of each TC positions and from 1 day before to 30 days after the TC passage.
19
Masked regions are then filled using bilinear interpolation from neighbouring regions. As a result, these 20
forced simulations do not account for any ocean feedback. Comparing TCs in the coupled and forced 21
simulations will therefore allow us to infer the impact of air-sea coupling on TC characteristics in the IO.
22
An illustration of the strategy employed to remove the SST signature of TCs in the forced experiment is 23
provided on Figure 2: as can be seen on that particular example, the cold wake is efficiently supressed from 24
the KF-FOR surface boundary condition.
25 26
The small perturbations induced by the cold wake removal are sufficient to change the courses of TCs
27
between the forced and coupled simulations, due to the chaotic nature of the atmosphere. Trajectories of 1
simulated TCs in the forced and coupled experiments will therefore be different and TCs in these two 2
simulations cannot be compared individually. Comparing the statistics of the simulated TCs in the forced 3
and coupled experiments will however allow us to infer the influence of air-sea coupling on IO TC 4
characteristics. A similar strategy was successfully used by Jullien et al. (2014) to assess the impact of air- 5
sea feedback under South Pacific TCs.
6 7
2.3. Tracking methodology and cyclogenesis indices
8 9
TCs from both simulations are tracked using the same methodology as in Samson et al. (2014). The 10
following criteria are used to distinguish tropical cyclones from intense mid-latitude systems at each time 11
step:
12
10 m wind > 17.5 m.s
-1associated with a local sea level pressure minimum 13
850 hPa vorticity > 3*10
-4s
-114
700–300 hPa mean temperature anomaly > 1°K 15
TC temperature anomalies are calculated with respect to their large-scale environment: the TC region is 16
defined as 3 radii of maximum wind around the TC centre while the environmental temperature is averaged 17
between 6 and 9 radii. Trajectories are then constructed by recursively detecting the closest neighbouring 18
grid points that meet all above criteria. If no matching point is identified, all criteria are relaxed except 19
vorticity. This relaxation technique allows following TCs over land and avoids counting the same TC 20
twice. Tracks shorter than 1 day are eliminated. The vorticity and temperature thresholds are similar to 21
those considered in previous studies (Jourdain et al. 2011; Jullien et al. 2014; Samson et al. 2014).
22 23
In the rest of the paper, we define TCs as storms verifying the above criteria. TC definition from official 24
centres can differ from one TC-prone basin to another, rendering comparison of TCs characteristics 25
between basins difficult: for instance, a TC is defined as a tropical system with sustained 10-m winds of at 26
least 17 m.s
-1in the west Australian and the northern IO basins, while this threshold is 33 m.s
-1for the
27
southwestern IO. Following the TCs official definition from the World Meteorological Organization, we 1
define TCs using a 17.5 m-s
-1on the sustained 10-m winds and use the Saffir-Simpson scale to define TC 2
categories as: tropical storm (Cat-0 TC for 17.5 m.s
-1<V
max<33 m.s
-1), Category 1 TC (Cat-1 TC for 33 m.s
-3
1
